Learning decision policies from confounded observational data is a challenging task in causal inference, as unobserved confounders can lead to biased or suboptimal actions when relying solely on machine learning models. A synergistic approach is learning to defer, which decides when to act itself and when to defer to a human expert with access to unobserved information. However, constructing the learning target, which defines the probability of choosing each action or deferral, remains a core challenge. To address this, we propose causal-target-based learning to defer (CTLD) framework, where the causal target is constructed from sharp bounds on potential outcomes. Specifically, the degree of overlap between these bounds determines the probability of deferral, while their relative positions and widths define the probabilities over actions. CTLD aligns model predictions with this causal target to make probabilistic decisions over actions and deferral. We present comprehensive theoretical guarantees for the learned policy and demonstrate the effectiveness of CTLD on synthetic and semi-synthetic datasets.

Using Web Mining and Content Analysis to find and understand clues from the massive amount of unstructured text online is very important for predicting future events and providing early risk warnings in important fields like finance and public safety. While Large Language Models (LLMs) exhibit potential in processing and understanding text, current text-based event prediction faces two primary challenges: first, an insufficient utilization of potential information within the text, such as causal relationships and latent associations, and second, limited predictive reliability constrained by issues like the LLM's own ability and hallucinations. To address these challenges, we propose a novel event prediction framework, Bidirectional Reasoning with Self-Correction (BRSC). BRSC comprises two complementary reasoning dimensions: temporal deductive reasoning, which analyzes the trajectory of historical events along the timeline to enable accurate trend extrapolation, and synchronic associative reasoning, which deeply mines details and latent connections from documents within a specific time window to extended semantic information. In addition, we use a self-correction mechanism that identifies and rectifies potential hallucinations and errors during the reasoning process. Extensive experiments on international relations event prediction demonstrate that BRSC achieves significant improvements over several leading LLM-based methods.

Large language models (LLMs) demonstrate remarkable capabilities in understanding complex tasks and have achieved commendable performance in graph-related tasks, such as node classification, link prediction, and subgraph classification.These tasks primarily depend on the local reasoning capabilities of the graph structure.However, research has yet to address the graph partitioning task that requires global perception abilities.Our preliminary findings reveal that vanilla LLMs can only handle graph partitioning on extremely small-scale graphs.To overcome this limitation, we propose a three-phase pipeline to empower LLMs for large-scale graph partitioning: coarsening, reasoning, and refining.The coarsening phase reduces graph complexity.The reasoning phase captures both global and local patterns to generate a coarse partition.The refining phase ensures topological consistency by projecting the coarse-grained partitioning results back to the original graph structure.Extensive experiments demonstrate that our framework enables LLMs to perform graph partitioning across varying graph scales, validating both the effectiveness of LLMs for partitioning tasks and the practical utility of our proposed methodology.

Humanitarian relief operations are often accompanied by regional conflicts around the globe, at risk of deliberate, persistent and unpredictable attacks. However, the long-term channeling of aid resources into conflict areas may influence subsequent patterns of violence and expose local communities to new risks. In this paper, we quantitatively analyze the potential externalities associated with long-lasting humanitarian relief operations based on game-theoretical modeling and online planning approaches. Specifically, we first model the problem of long-lasting humanitarian relief operations in conflict areas as an online multi-stage rescuer-and-attacker interdiction game in which aid demands are revealed in an online fashion. Both models of single-source and multiple-source relief supply policy are established respectively, and two corresponding near-optimal online algorithms are proposed. In conjunction with a real case of anti-Ebola practice in conflict areas of DR Congo, we find that 1) long-lasting humanitarian relief operations aiming alleviation of crises in conflict areas can lead to indirect funding of local rebel groups; 2) the operations can activate the rebel groups to some extent, as evidenced by the scope expansion of their activities. Furthermore, the impacts of humanitarian aid intensity, frequency and supply policies on the above externalities are quantitatively analyzed, which will provide enlightening decision-making support for the implementation of related operations in the future.

Maintaining and upgrading underground pipelines are major undertakings in urban operations, where accurately locating buried pipelines has long been an issue. In this article, we propose a pipeline mapping method based on integrating multipositional pipeline data, which includes the multisensor data acquisition (MDA) platform and the scalable probability-based pipeline mapping (SP-PM) model. To effectively collect pipeline data at multiple positions, several pipeline detecting and positioning sensors are equipped in the MDA platform. Different types of sensor data are synchronously collected and processed to obtain manageable pipeline data at multiple positions within the detected area, such as the radius, depth, and positioned points of underground pipelines. The SP-PM model is then proposed, where the obtained pipeline data are probabilistically described and classified into classes. Each class contains the pipeline data possibly generated by the same pipeline. The classified data are then iteratively integrated to estimate the pipeline map with the maximum probability. The SP-PM model probabilistically estimates the degree of correlation between the multipositional pipeline data and each potential pipeline, and it has no strict requirement on existing statutory records or limits on the number of detections per pipeline. Unrecorded pipelines could be identified and involved into the generated pipeline map, along with continuous adjusting of the pipelines’ number. We conducted experiments on real-world environments. The experimental results verify the accuracy and efficiency of the proposed method for buried pipeline mapping.

Wireless sensor networks (WSNs) are an effective and efficient method for collecting data from a target area, and prolonging the lifetime of WSNs has been a focus of scientific research due to the limited energy of sensor nodes. However, traditional studies of WSNs are based on the assumption that WSNs collect aggregated data with redundant sensor nodes in an ideal radio environment. These assumptions may not be acceptable in practical applications. To address this problem, this paper introduces a novel application scenario for WSNs in which raw data are collected by a multihop network without redundant sensor nodes, and a new hybrid tree-based and cluster-based routing protocol for raw data collection (HTC-RDC) is proposed to prolong the lifetime of WSNs that can work in novel application scenarios. The experimental results demonstrate that the proposed HTC-RDC enables the WSNs to achieve their expected functions and prolongs the network lifetime by an average of 11.4% compared to the existing protocols in the explored application scenario.

Participatory sensing is a crowdsourcing-based framework, where the platform executes the sensing requests with the help of many common peoples' handheld devices (typically smartphones). In this paper, we mainly address the online sensing request admission and smartphone selection problem to maximize the profit of the platform, taking into account the queue backlog, and the location of sensing requests and smartphones. First, we formulate this problem as a discrete time model and design a location aware online admission and selection control algorithm (LAAS) based on the Lyapunov optimization technique. The LAAS algorithm only depends on the currently available information and makes all the control decisions independently and simultaneously. Next, we utilize the recent advancement of the accurate prediction of smartphones' mobility and sensing request arrival information in the next few time slots and develop a predictive location aware admission and selection control algorithm (PLAAS). We further design a greedy predictive location aware admission and selection control algorithm (GPLAAS) to achieve the online implementation of PLAAS approximately and iteratively. Theoretical analysis shows that under any control parameter V > 0, both LAAS and PLAAS algorithm can achieve O(1/V)-optimal average profit, while the sensing request backlog is bounded by O(V). Extensive numerical results based on both synthetic and real trace show that LAAS outperforms the Greedy algorithm and Random algorithm and GPLAAS improves the profit-backlog tradeoff over LAAS.